Evolution of Hollow TiO 2 Nanostructures via the Kirkendall Effect Driven by Cation Exchange with Enhanced Photoelectrochemical Performance

Transcription

1 pubs.acs.org/nanolett Evolution of Hollow TiO 2 Nanostructures via the Kirkendall Effect Driven by Cation Exchange with Enhanced Photoelectrochemical Performance Yanhao Yu, Xin Yin, Alexander Kvit,, and Xudong Wang*, Department of Materials Science and Engineering and Materials Science Center, University of Wisconsin-Madison, Madison, Wisconsin 53706, United States *S Supporting Information ABSTRACT: Hollow nanostructures are promising building blocks for electrode scaffolds and catalyst carriers in energyrelated systems. In this paper, we report a discovery of hollow TiO 2 nanostructure evolution in a vapor solid deposition system. By introducing TiCl 4 vapor pulses to ZnO nanowire templates, we obtained TiO 2 tubular nanostructures with wellpreserved dimensions and morphology. This process involved the cation exchange reaction between TiCl 4 vapor and ZnO solid and the diffusion of reactants and products in their vapor or solid phases, which was likely a manifestation of the Kirkendall effect. The characteristic morphologies and the evolution phenomena of the hollow nanostructures from this vapor solid system were in a good agreement with the Kirkendall effect discovered in solution systems. Complex hollow TiO 2 nanostructures were successfully acquired by replicating various ZnO nanomorphologies, suggesting that this unique cation exchange process could also be a versatile tool for nanostructure replication in vapor solid growth systems. The evolution of TiO 2 nanotubes from ZnO NW scaffolds was seamlessly integrated with TiO 2 NR branch growth and thus realized a pure TiO 2 -phased 3D NW architecture. Because of the significantly enlarged surface area and the trace amount of Zn left in the TiO 2 crystals, such 3D TiO 2 nanoforests demonstrated enhanced photoelectrochemical performance particularly under AM (air mass) 1.5G illumination, offering a new route for hierarchical functional nanomaterial assembly and application. KEYWORDS: Hollow TiO 2 nanostructure, vapor solid deposition, cation exchange, Kirkendall effect, photoelectrochemical water splitting Mesoscale morphology engineering is an effective approach to tailoring material s physical properties, such as light interaction, surface reaction, and charge transport for achieving enhanced performance. Hollow structures with mesoscale pores are one typical example that shows attractive morphology-enabled or enhanced performance in the areas of energy harvesting and storage, catalysts design, and biomedicine development. 1 3 Among various approaches, the Kirkendall effect was found fairly effective in producing hollow nanostructures from a broad range of materials, including metal chalcogenides, 4 12 metal oxides, metal phosphines, alloys and compound nanocrystals In a typical Kirkendall process, different ionic diffusion rates of the core and shell materials generate a net inward vacancy diffusion, which leads to voids nucleation at the core/shell interface and eventually the formation of a hollow structure. Solid bridges were usually found existing between the core and shell to facilitate the transport of core materials outward The Kirkendall effect is a unique self-templating strategy in synthesizing nanostructures with large yield, high crystal quality, and uniform size and shape. Currently, investigations and applications of the Kirkendall effect in nanomaterial syntheses are mostly focused on nanoparticle processing in solution systems. Solid-state Kirkendall effect was also observed in 1D hollow nanomaterial (e.g., nanotubes) evolution, whereas the reaction kinetics were much slower due to the limited diffusivities. 29,31,32 In this paper, we report a discovery of mesoscale hollow TiO 2 nanostructure evolution via cation exchange reactions, which was likely a manifestation of the Kirkendall effect in a vapor solid deposition system. This process involved the cation exchange reaction between TiCl 4 vapor and ZnO solid forming TiO 2 and the diffusion of the reactants and products through the TiO 2 shell. Different from typical cation exchange results in nanostructure replication, where solid morphology is always received, we obtained complex hollow TiO 2 nanostructures with a well-preserved morphology. Further integrating this process with the surface-reaction-limited pulsed chemical vapor deposition (SPCVD) technique successfully Received: January 24, 2014 Revised: March 14, 2014 Published: March 28, American Chemical Society 2528

2 Nano s Figure 1. Evolution of TiO 2 nanotubes. (a) SEM image of TiO 2 nanotubes synthesized via 10 cycles of TiCl 4 and H 2 O deposition. Inset is SEM image of the ZnO NWs template, where the similar morphology to TiO 2 nanotubes can be identified. Scale bar is 200 nm. (b) XRD patterns of TiO 2 samples collected from a series of deposition cycles. (c e) Low- and high-magnification TEM images of TiO 2 samples collected from 2 cycles (c), 5 cycles, (d) and 10 cycles (e) of deposition. Insets in c-ii e-ii are corresponding cross-sectional SEM images (scale bars are 20 nm). Inset in e-i is SAED image taken from the nanotube shown in (e). created high density three-dimensional (3D) TiO 2 nanowire (NW) architecture. Trace amount Zn left in the TiO 2 shell significantly improved its photoelectrochemical (PEC) performance. This discovery opens a new route toward creating complex hollow nanostructures. A homemade atomic layer deposition (ALD) system was adapted to provide pulsed TiCl 4 and H 2 O vapor to ZnO NW templates (inset of Figure 1) at 600 C for structure replication. The growth followed typical ALD operation cycles where TiCl 4 and H 2 O pulses were separated by 60 s N 2 purges (see the Supporting Information for experiment details). The ZnO NW structures were well preserved after 10 cycles of deposition as shown in Figure 1a, whereas the surfaces became rough. The surface roughening began to be observable after 2 cycles of growth (Supporting Information Figure S1). X-ray diffraction (XRD) spectra acquired at different deposition cycles revealed the phase evolution during the growth (Figure 1b). Upon the introduction of TiCl 4 and H 2 O precursors, the ZnO phase decreased and completely disappeared from the samples after 10 cycles of deposition. Meanwhile, anatase TiO 2 accumulated continuously with the increase of deposition cycles, indicating the ZnO NWs has been converted to TiO 2 during the process. Transmission electron microscopy (TEM) shows that the asobtained 1D TiO 2 nanostructure was polycrystalline nanotubes. Figure 1c e shows a series of TEM images revealing the evolution of TiO 2 nanotubes. A polycrystalline layer of 8 nm was observed on the surface of ZnO NW after 2 cycles of deposition. This thickness was significantly larger than a typical ALD TiO 2 film ( 0.1 nm per cycle), suggesting chemical vapor reaction was playing a dominating role during the precursor exposure. Voids can be clearly identified at the core/ shell interface from the 2-cycle sample by the distinct contrast in TEM images and the SEM cross-section image (Figure 1c-ii and inset). From the 5-cycle sample, disordered and intense contrast of crystalline grains were observed by TEM (Figure 1d), which was a result of the rapidly grown polycrystalline TiO 2 shell and the random core residue. Although not observable in TEM images, voids and bridges can be recognized from the cross-sectional SEM image (inset of Figure 1d-ii). Consistent with the XRD observation, pure-phased polycrystalline anatase TiO 2 nanotubes with a shell thickness of nm were received after 10 cycles of deposition, as shown by the TEM image (Figure 1e) and corresponding selective area electron diffraction (SAED) pattern (inset of Figure 1e). The typical TiO 2 grain size was in the range of nm (Figure 1e-ii). The average inner diameter of the as-received TiO 2 nanotubes was measured to be 40 nm, which was 67% of the original ZnO NW sizes (averaging 60 nm in diameter). This ratio is within the size ratio range (40 70%) reported for hollow nanoparticles evolved via the Kirkendall effect. 30 This size fact, together with the observation of interface voids and solid bridges, and the rapid shell growth, suggests that the hollow TiO 2 nanostructures were very likely formed following the Kirkendall effect. It is also important to point out that the direct reaction between TiCl 4 vapor and ZnO NW is a cation exchange reaction TiCl4 + 2ZnO TiO2 + 2ZnCl2 (1) This is likely to be the mechanism of the formation of the initial TiO 2 layer on the ZnO surfaces. 41 However, when a complete TiO 2 shell is formed, solid-state diffusion has to take place in 2529

3 Nano s Figure 2. Microscopic study of TiO 2 nanotube evolution. (a) STEM image and EELS spectrum imaging of the Ti, Zn, and O elements on a 2-cycle sample. The ZnO core residue is highlighted by a white dashed box. (b) TEM images of a TiO 2 nanotube formed via 10 cycles of TiCl 4 pulse without introducing any H 2 O precursor. Inset on b-i is the corresponding cross-sectional SEM image showing the hollow feature; (b-ii) highermagnification TEM images; (c-i,ii) TEM images of amorphous TiO 2 coated ZnO NW after 10 h annealing. Voids at the core/shell interface are marked with white dashed lines. Inset on c-i is the corresponding cross-sectional SEM image. (d) Schematic illustration of the Kirkendall effect combined with cation exchange process in TiO 2 nanotube evolution. order to form a center-hollowed structure, which is beyond the operational range of the cation exchange mechanism. In typical cation exchange processes, it is generally accepted that cations are replaced in original structure while anionic framework remains intact and thus the size and shape of the original crystal templates are usually well-conserved However, it should be noted that the different valences of zinc and titanium ions (one Ti 4+ replaces two Zn 2+ ) might lead to voids formation as well. Although the misvalence cation exchange has not been clearly discovered in oxide systems, research in sulfides and selenides showed consistent results that no hollow structure can be created through the mis-valence cation exchange. For example, cation exchange pairs including CdSe/Ag 2 Se, 33 PbS/ Cu 2 S, 36 and ZnSe/Cu 2 Se 37 were investigated for nanostructure replication, where all of them preserved the original size and shape and no voids were observed. On the basis of above rationale, it becomes conceivable that the cation exchange alone is not sufficient to form the TiO 2 hollow nanostructures. The Kirkendall effect should have played a critical role in the hollow structure evolution. To further confirm the fact of Zn element diffusion during TiO 2 shell evolution, electron energy loss spectroscopy (EELS) mapping was conducted on the 2-cycle sample in scanning transmission electron microscopy (STEM) (Figure 2a). Ti distribution clearly matched the grain morphology along the polycrystalline shell. The Zn map revealed a universal distribution of Zn in the entire structure with strong Zn signals located at the center region representing the remaining ZnO core (marked by the dashed white box). The homogeneous zinc distribution along the entire shell region suggests that the Zn elements from the core were diffusing into and traveling through the TiO 2 shells during the growth. Because no new phase except ZnO and TiO 2 were observed by XRD, it was likely that Zn elements only diffused through the TiO 2 shell but did not form any new composite. In order to understand the chemical reaction mechanism, H 2 O or TiCl 4 precursors were removed individually during the pulsed deposition while all the other reaction conditions were kept the same. When only TiCl 4 vapor was present, 10-cycle deposition also completely removed the ZnO core and same anatase TiO 2 (see Supporting Information Figure S2a for XRD phase confirmation) polycrystalline nanotubes were obtained, as shown by the SEM and TEM images in Figure 2b. When H 2 O vapor precursor was used exclusively, only small pores were created on the ZnO NW surface (Supporting Information Figure S3a,b). This comparison reveals that H 2 O was an insignificant component in the ZnO to TiO 2 conversion, and the O in TiO 2 can be completely provided by ZnO following the cation exchange reaction (eq 1). The low-melting temperature (292 C) and high vapor pressure (1 Torr at 408 C) of ZnCl 2 suggest that the ZnCl 2 byproducts would most likely exist in its vapor phase, which was favorable for rapid diffusion and complete removal of the Zn species from the core. When the deposition temperature was dropped to 300 C under the exposure of only TiCl 4, the diffusion of Zn species was significantly limited and a ZnO/TiO 2 core/shell structure 2530

4 Nano s Figure 3. Microscopic characterization of TiO 2 replication of plate-caped ZnO NW bundles by 300 cycles (a) and 600 cycles (b) of deposition. (i) SEM images showing the morphology of ZnO template was uniformly and precisely replicated after 300 and 600 deposition. Inset of a-i is an image of a broken region showing the core residue in the plate cap and the hollow center in the NW stem. Inset of b-i shows a hollow center in the plate region. (ii) TEM images of individual hollow nanostructures, and the detailed structure features in the cap and NW regions are shown in (iii) and (iv), respectively. Inset of a-ii is the corresponding SAED pattern confirming the anatase phase. The dark contrast in the cap region on a-iii and b-iii corresponds to the ZnO residue, which is located near to the interior shell surface. (v) STEM image and EDS spectrum imaging of Ti, Zn, and O on the replicated TiO 2 hollow nanostructures. Obvious Zn residue concentrated along the interior surface of shell, while a uniform and low Zn signal can be identified over the entire nanostructure evidencing the existence of trace amount of Zn in the TiO 2 shells. was received (Supporting Information Figure S3c,d). This result represented the cation exchange reaction and revealed the critical role of high-temperature diffusion in the implementation of the Kirkendall effect. Further evidence of the critical role of TiCl 4 in the formation of pure-phased TiO 2 nanotubes came from annealing of amorphous TiO 2 coated ZnO NWs. Amorphous TiO 2 layer with a thickness of 14 nm was first deposited by ALD at 100 C and then the temperature was immediately raised to 600 C to anneal the sample for 10 h in N 2 atmosphere. This process preserved a significant amount of ZnO cores; while multiple voids were observed at the core/shell interface (Figure 2c). The shell was found to be ZnTiO 3 (Supporting Information Figure S2b), which could be a result of the solid-state reaction between ZnO and TiO 2. This observation agrees with the previously reported solid-state Kirekendall effect in ZnAl 2 O 4 spinel nanotube formation. 29 This experiment revealed that exclusive solid-state diffusion of ZnO species cannot adequately explain the vanishing of ZnO in the heterostructure and formation of pure TiO 2 phase. The above experiments confirmed that TiCl 4 exposure is necessary for the formation of pure-phase TiO 2 nanotubes, however we found it is not sufficient. When the same amount of TiCl 4 exposure was introduced to the ZnO NWs without the purging intervals (5 s continuous TiCl 4 pulse = 10 cycles of 0.5 s TiCl 4 pulse), a core/shell structure was obtained (Supporting Information Figure S4). The shell appeared to be polycrystalline TiO 2 with a thickness of 17 nm and smaller grain sizes ( 5 nm). A large portion of the ZnO NW core was preserved. This result revealed that it is important to leave a sufficiently long dwelling time between TiCl 4 pulses (i.e., the purging step in ALD processes) allowing the core species to diffuse through the solid shell and react with the gas phase reactants for the formation of a complete hollow structure. On the basis of our experimental observations, we suggest that the formation of hollow TiO 2 nanostructures likely follows the Kirkendall effect driven by cation exchange reactions, and the proposed mechanism is schematically illustrated in Figure 2d. When TiCl 4 vapor is first introduced at 600 C, the exterior surface of the ZnO NW quickly converts to TiO 2 via cation exchange and generates ZnCl 2 byproduct most likely in a vapor phase. Meanwhile, initial voids were created at the defective core/shell interface. Once the surface is fully covered by the TiO 2 shell, further removal of ZnO required the Zn species to diffuse outward. This process could be explained by inward diffusion of the TiCl 4 vapor and outward diffusion of Zn 2+ O 2 through the polycrystalline TiO 2 shell. (It is likely that both Zn 2+ and O 2 are diffused in their ionic form. For simplicity, we just use ZnO to represent this case here.) The cation exchange reaction between TiCl 4 and ZnO occurs at the TiO 2 /ZnO interface and TiO 2 outer surface, resulting inward and outward growth of the shell, respectively. This reaction may also occur inside the shell and fill the spaces between TiO 2 grains. The byproduct, ZnCl 2 vapor is able to diffuse through the TiO 2 shell and be removed through pumping. Upon the consumption of ZnO at the TiO 2 /ZnO interface, the void size increases and ZnO transports from the NW center to the interface area leaving a typical Kirkendall bridge morphology. Eventually, the entire ZnO core is consumed yielding a hollow TiO 2 nanotube. The complete nanotube evolution takes less than 30 min, which is comparable to nanoscale Kirkendall evolution in solution 14,30 and much faster than solid-state reactions 29,31 at same size scale. 2531

5 Nano s Figure 4. Development of TiO 2 3D NW architecture (nanoforest). (a) Cross-sectional SEM image of a 3D TiO 2 nanoforest formed via 400-cycle SPCVD growth. Dense TiO 2 NRs were deposited uniformly along the entire TiO 2 nanotube trunk. Inset is the top view SEM image. (b) TEM image of an individual branched TiO 2 NW. Inset is the corresponding SAED pattern confirming the anatase phase. (c) Higher-magnification TEM image shows the single-crystalline NR branch and the polycrystalline nanotube trunk. (d) XRD pattern of the overall TiO 2 nanoforest sample identifies both anantase and rutile TiO 2 phases, however no ZnO peaks can be identified. Discovery of the unique evolution process of hollow TiO 2 nanostructures in a vapor solid system offered an effective approach for fabricating TiO 2 nanostructures using the splendid nanomorphologies of ZnO as the template. 42 To illustrate the unique shape preservation capability, ZnO plate-capped NW bundles were selected as the template for the pulsed TiCl 4 CVD process (Supporting Information Figure S5a,b). Because of the large size of the plate caps ( 1 μm), more cycles of deposition were applied for structure replication. After 300 cycles of deposition, the plate-capped NW morphology was successfully replicated by TiO 2 with a hollow feature and slightly rough surfaces (Figure 3a-i). ZnO in the NW portion ( 50 nm in diameter) was completely converted into a hollow nanotube structure (inset of Figure 3a-i). However, in the large plate region, considerable amount of ZnO still remained near the interior surface of the shell, as highlighted by the white dashed line in the inset of Figure 3a-i. The hollow structure was further confirmed by TEM and annular bright-field STEM characterization (Figure 3a-ii and Supporting Information Figure S5c), where the darker contrast near the interior surface of the cap region may correspond to the core residue as observed by SEM (Figure 3a-iii). The NW was completely converted into a nanotube with a wall thickness of 50 nm (Figure 3a-iv). The pure polycrystalline anatase TiO 2 phase was confirmed by the SAED pattern that is shown in the inset of Figure 3a-ii. Energy dispersive X-ray spectroscopy (EDS) spectrum imaging acquired on the cap region also confirms the distribution of concentrated zinc residue along the interior surface of the shell (Figure 3a-v). By further increasing the replication process to 600 cycles, most ZnO residues were removed leaving a nearly complete hollow TiO 2 nanostructure, while the overall morphology was 2532

6 Nano s still well-preserved. (Figure 3b-i,ii and Supporting Information Figure S5d). The dark contrast region became much smaller compared to the 300-cycle sample and only located at the tip where the ZnO core had the largest volume (Figure 3b-iii). Such a Zn distribution was also confirmed by EDS mapping (Figure 3b-v). The thickness of the nanotube was found increased to 80 nm (Figure 3b-iv), revealing that the additional 300 cycles of deposition resulted in 30 nm growth, corresponding to a growth rate of 0.1 nm/s. This value was very close to the typical ALD growth rate, suggesting that after the ZnO core was completely consumed, the growth would come back to regular ALD process. It was also interesting to see a low concentration but uniform Zn background in the entire TiO 2 replica area from the EDS mappings of both 300-cycle and 600- cycle samples (Figure 3a-v,b-v), suggesting the existence of trace amount of Zn elements in the TiO 2 shell, which may change the physical property of anatase TiO 2. To demonstrate the universality of this approach in morphology duplication, we also successfully converted a large-area 3D-interconnected ZnO nanosheet structure into a hollow TiO 2 nanostructure through 400 cycles of deposition (Supporting Information Figure S6). All the experimental results confirmed that in principle, this approach can be employed to replicate ZnO nanostructures with any morphology via a vapor solid process. This strategy may be further adapted to the replication of other materials if appropriate chemical reactions can be identified. The TiO 2 nanotube arrays could serve as an excellent scaffold to support dense TiO 2 NR branch growth via our recently developed SPCVD technique, yielding a pure TiO 2 -phased 3D NW architecture (Figure 4a). Using ZnO NW arrays as the initial template, 400 total cycles of TiCl 4 and H 2 O precursor exposures at 600 C deposited high-density TiO 2 NR branches uniformly covering all TiO 2 nanotubes that were converted from the ZnO NWs (Supporting Information Figure S7). TEM characterization on a single hierarchical NW structure further revealed the dense single-crystalline TiO 2 NR coverage (Figure 4b) and the polycrystalline hollow trunk (Figure 4c). The SAED pattern taken from the hierarchical NW body confirmed the anatase phase (inset of Figure 4b). No ZnO phase was observed from either SAED or XRD pattern (Figure 4d), and the Zn signal was not detectable by EDS (Supporting Information Figure S7), which further confirmed the hollow feature. From the XRD pattern, rutile TiO 2 was also observed, which might be attributed to the lower anatase phase stability of the polycrystalline trunk when foreign element (e.g., Zn) was involved during the formation. 46 Before this work, TiO 2 branches could not be synthesized by the SPCVD technique on either single- or poly crystalline TiO 2 cores, because of the higher free energy of nucleating new TiO 2 seeds compared to that of epitaxial film growth. 47 Here, the TiO 2 NR branch growth seamlessly followed the NW-tonanotube conversion under the same conditions, and the resulted TiO 2 crystals were favorable for NR SPCVD growth. A series of SEM images taken from samples collected at different deposition cycles revealed this evolution process (Supporting Information Figure S8). Anisotropic crystal growth became obvious from the 25-cycle sample and continued to enlarge the crystals aspect ratio and eventually formed the NR morphology. This observation was in a good agreement with the TiO 2 NR evolution observed on Si surfaces. 47 Therefore, it can be concluded that the cation exchange and Kirkendall effect were only in response for the initial removal of ZnO core and 2533 forming TiO 2 nanotubes, after which the crystals would switch back to their regular growth behavior the NR growth in this case. The 3D hierarchical NW structure is believed to be a promising configuration for building high-performance PEC cells. 43,45,48 51 PEC water-splitting experiments were conducted to explore the application potential of the 3D TiO 2 nanoforest. In order to provide a performance comparison, two reference samples were fabricated by ALD coating a polycrystalline thin film of anatase TiO 2 on the same ZnO NW arrays ( 2 μm long) at 300 C. The TiO 2 coatings were 27 and 51 nm thick from 400 and 800 cycles of ALD, respectively (Supporting Information Figure S9), corresponding to equal and twice amount of TiO 2 exposure as that for TiO 2 NR growth. The ZnO NW core was not attacked during the deposition (Supporting Information Figure S9). Figure 5 shows the Figure 5. PEC performance of 3D TiO 2 nanoforest. (a,b) Photocurrent density (solid line) versus bias potential (versus SCE) and dark current (dashed line) of measured from photoanodes made from the TiO 2 nanoforest (blue), and ZnO NW arrays coated with TiO 2 films deposited by 400 cycles (red) and 800 cycles (yellow) ALD. The characterization was conducted under the illumination of 150 W xenon lamp (a); and AM 1.5G with a light intensity of 100 mw/cm 2 (b). photocurrent density versus applied bias acquired from the 3D TiO 2 nanoforest and the two reference samples under the illumination of a 150 W Xe lamp (a) and 100 mw/cm 2 AM 1.5G (b). Under direct Xe lamp illumination, three samples exhibited similar open circuit voltage (V OC ) and followed the same trend; whereas, their short circuit photocurrent densities (J SC ) were dramatically different. The two reference samples exhibited similar J SC ( 1.86 ma/cm 2 for 27 nm TiO 2 coating

7 Nano s and 1.74 ma/cm 2 for 51 nm TiO 2 coating), indicating that simply increasing the amount of TiO 2 was not helpful to the photocurrent improvement because of the surface-dominated oxidation reactions. Owing to the enlarged surface area, the 3D TiO 2 nanoforest achieved a J SC of 3.14 ma/cm 2, which was remarkably higher than other reported TiO 2 NW-based PEC electrodes. 40,43 The J SC difference became more conspicuous under AM 1.5G illumination, although the absolute values were much smaller than those obtained under Xe lamp due to the cut off of UV light. As shown in Figure 5b, J SC of the 3D TiO 2 nanoforest ( 0.75 ma/cm 2 ) was 7 times higher than both reference samples ( 0.11 ma/cm 2 ). This observation suggests that the 3D TiO 2 nanoforest had better photo response in the visible light regime. This enhancement may be attributed to the trace amount of Zn element left in the TiO 2 crystals, which might change the absorption or charge transport properties (note that in the TiO 2 nanoforest structure, all the charges were conducted through the polycrystalline TiO 2 nanotube trunk). In addition, the slightly larger V OC and higher dark current of 3D TiO 2 nanoforest also suggested the modified electronic properties probably caused by the Zn impurities. In summary, various hollow TiO 2 nanostructures were achieved in the pulsed vapor deposition system when TiCl 4 vapor was interacting with ZnO NWs with extended reaction time. Through this process, ZnO NWs were quickly and completely converted into polycrystalline TiO 2 hollow nanostructures, while the NW dimension and morphology were well preserved. Trace amount of Zn elements were found left in the TiO 2 crystals, which may change the electronic properties of the TiO 2 nanostructures. The characteristic morphologies and the evolution phenomena of the hollow nanostructures from this vapor solid system were in a good agreement with the Kirkendall effect discovered in solution systems. Similarly, this approach is not limited by the nanomaterial s morphology in the vapor solid process. Complex hollow TiO 2 nanostructures were successfully acquired by replicating various ZnO nanostructures, including plate-caped NWs and 3D nanosheet networks. This work suggested that combining cation exchange reactions with the Kirkendall effect might be a versatile tool for creating complex hollow nanostructures from the vapor phase deposition. On the basis of the principle observed from the TiO 2 ZnO system, we believe this strategy can be applied to other materials if appropriate chemical reactions between the vapor reactant and the solid template can be identified. We also found that after the template was completely converted, further deposition would switch back to its regular growth pattern. Therefore, the formation of TiO 2 nanotubes from ZnO NW scaffold was seamlessly integrated with TiO 2 NR branch growth and thus a pure TiO 2 -phased 3D NW architecture was achieved. Such 3D TiO 2 nanoforests demonstrated enhanced PEC performance particularly under AM 1.5G illumination. The discovery of automatic growth behavior switch would open a new route for hierarchical nanomaterial assembly. ASSOCIATED CONTENT *S Supporting Information Experimental details and SEM images, XRD patterns, TEM images, and STEM images. This material is available free of charge via the Internet at AUTHOR INFORMATION Corresponding Author * xudong@engr.wisc.edu. Notes The authors declare no competing financial interest. ACKNOWLEDGMENTS Research primarily supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES), under Award # DE-SC , and Air Force Office of Scientific Research (AFOSR) under Award # FA REFERENCES (1) Lou, X. W.; Archer, L. A.; Yang, Z. Adv. Mater. 2008, 20, (2) Wang, W.; Dahl, M.; Yin, Y. Chem. Mater. 2013, 25, (3) Fan, H. J.; Gosele, U.; Zacharias, M. Small 2007, 3, (4) Yin, Y.; Erdonmez, C. K.; Cabot, A.; Hughes, S.; Alivisatos, A. P. Adv. Funct. Mater. 2006, 16, (5) Tang, Y.; Ouyang, M. Nat. Mater. 2007, 6, 754. (6) Cabot, A.; Ibanez, M.; Guardia, P.; Alivisatos, A. P. J. Am. Chem. Soc. 2009, 131, (7) Gao, J.; Zhang, B.; Zhang, X.; Xu, B. Angew. Chem., Int. Ed. 2006, 45, (8) Wang, Y.; Cai, L.; Xia, Y. Adv. Mater. 2005, 17, 473. (9) Tan, H.; Li, S.; Fan, W. Y. J. Phys. Chem. B 2006, 110, (10) Cabot, A.; Smith, R. K.; Yin, Y.; Zheng, H.; Reinhard, B. M.; Liu, H.; Alivisatos, A. P. ACS Nano 2008, 2, (11) Wang, W.; Goebl, J.; He, L.; Aloni, S.; Hu, Y.; Zhen, L.; Yin, Y. J. Am. Chem. Soc. 2010, 132, (12) Zhang, G.; Wang, W.; Yu, Q.; Li, X. Chem. Mater. 2009, 21, 969. (13) Cabot, A.; Puntes, V. F.; Shevchenko, E.; Yin, Y.; Balcells, L.; Marcus, M. A.; Hughes, S. M.; Alivisatos, A. P. J. Am. Chem. Soc. 2007, 129, (14) Niu, K. Y.; Park, J.; Zheng, H.; Alivisatos, A. P. Nano Lett. 2013, 13, (15) Peng, S.; Sun, S. Angew. Chem., Int. Ed. 2007, 46, (16) Railsback, J. G.; Johnston-Peck, A. C.; Wang, J.; Tracy, J. B. ACS Nano 2010, 4, (17) Nakamura, R.; Tokozakura, D.; Nakajima, H.; Lee, J. G.; Mori, H. J. Appl. Phys. 2007, 101, (18) Chiang, R. K.; Chiang, R. T. Inorg. Chem. 2007, 46, 369. (19) Henkes, A. E.; Vasquez, Y.; Schaak, R. E. J. Am. Chem. Soc. 2007, 129, (20) Wang, J.; Johnston-Peck, A. C.; Tracy, J. B. Chem. Mater. 2009, 21, (21) Sun, Y.; Xia, Y. Science 2002, 298, (22) Gonzalez, E.; Arbiol, J.; Puntes, V. F. Science 2011, 334, (23) Jana, S.; Chang, J. W.; Rioux, R. M. Nano Lett. 2013, 13, (24) Zhang, F.; Shi, Y.; Sun, X.; Zhao, D.; Stucky, G. D. Chem. Mater. 2009, 21, (25) Liang, X.; Wang, X.; Zhuang, Y.; Xu, B.; Kuang, S.; Li, Y. J. Am. Chem. Soc. 2008, 130, (26) Cao, H.; Qian, X.; Wang, C.; Ma, X.; Yin, J.; Zhu, Z. J. Am. Chem. Soc. 2005, 127, (27) Tian, L.; Yang, X.; Lu, P.; Williams, I. D.; Wang, C.; Ou, S.; Liang, C.; Wu, M. Inorg. Chem. 2008, 47, (28) Aldinger, F. Acta Met. 1974, 22, 923. (29) Fan, H. J.; Knez, M.; Scholz, R.; Hesse, D.; Nielsch, K.; Zacharias, M.; Gosele, U. Nano Lett. 2007, 7, 993. (30) Yin, Y.; Rioux, R. M.; Erdonmez, C. K.; Hughes, S.; Somorjai, G. A.; Alivisatos, A. P. Science 2004, 304, 711. (31) Fan, H. J.; Knez, M.; Scholz, R.; Nielsch, K.; Pippel, E.; Hesse, D.; Zacharias, M.; Gosele, U. Nat. Mater. 2006, 5, 627. (32) Peng, Q.; Sun, X. Y.; Spagnola, J. C.; Saquing, C.; Khan, S. A.; Spontak, R. J.; Parsons, G. N. ACS Nano 2009, 3, 546.

8 Nano s (33) Son, D. H.; Hughes, S. M.; Yin, Y.; Paul Alivisatos, A. Science 2004, 306, (34) Robinson, R. D.; Sadtler, B.; Demchenko, D. O.; Erdonmez, C. K.; Wang, L. W.; Alivisatos, A. P. Science 2007, 317, 355. (35) Rivest, J. B.; Jain, P. K. Chem. Soc. Rev. 2013, 42, 89. (36) Luther, J. M.; Zheng, H.; Sadtler, B.; Alivisatos, A. P. J. Am. Chem. Soc. 2009, 131, (37) Li, H.; Zanella, M.; Genovese, A.; Povia, M.; Falqui, A.; Giannini, C.; Manna, L. Nano Lett. 2011, 11, (38) Ritala, M.; Leskela, M.; Nyka nen, E.; Soininen, P.; Niinisto, L. Thin Solid Films 1993, 225, 288. (39) Law, M.; Greene, L. E.; Radenovic, A.; Kuykendall, T.; Liphardt, J.; Yang, P. J. Phys. Chem. B 2006, 110, (40) Hwang, Y. J.; Boukai, A.; Yang, P. Nano Lett. 2009, 9, 410. (41) Muduli, S.; Game, O.; Dhas, V.; Yengantiwar, A.; Ogale, S. B. Energy Environ. Sci. 2011, 4, (42) Wang, Z. L. J. Phys.: Condens. Matter 2004, 16, R829. (43) Shi, J.; Hara, Y.; Sun, C.; Anderson, M. A.; Wang, X. Nano Lett. 2011, 11, (44) Shi, J.; Sun, C.; Starr, M. B.; Wang, X. Nano Lett. 2011, 11, 624. (45) Shi, J.; Wang, X. Energy Environ. Sci. 2012, 5, (46) Shi, J.; Wang, X. Cryst. Growth Des. 2011, 11, 949. (47) Shi, J.; Li, Z.; Kvit, A.; Krylyuk, S.; Davydov, A. V.; Wang, X. Nano Lett. 2013, 13, (48) Liu, C.; Tang, J.; Chen, H. M.; Liu, B.; Yang, P. Nano Lett. 2013, 13, (49) Kargar, A.; Sun, K.; Jing, Y.; Choi, C.; Jeong, H.; Zhou, Y.; Madsen, K.; Naughton, P.; Jin, S.; Jung, G. Y.; Wang, D. Nano Lett. 2013, 13, (50) Cho, I. S.; Chen, Z.; Forman, A. J.; Kim, D. R.; Rao, P. M.; Jaramillo, T. F.; Zheng, X. Nano Lett. 2011, 11, (51) Kargar, A.; Sun, K.; Jing, Y.; Choi, C.; Jeong, H.; Jung, G. Y.; Jin, S.; Wang, D. ACS Nano 2013, 7,